Back to skill

Security audit

Bloomin8

Security checks for vulnerabilities and agentic risk

Overview

This skill appears purpose-built for Bloomin8 frame control, but it needs Review because it can send images and tokens over unauthenticated local HTTP and persistently reconfigure the device's upstream server.

Install only if you trust the local network and the Bloomin8 device you target. Avoid passing real upstream tokens on the command line, be cautious with custom upstream URLs, and review any action that scans Bluetooth, probes an IP address, uploads private images, or changes upstream settings.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Findings (3)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/bloomin8_cli.py:319
Finding

Plaintext and Unauthenticated Transport for Sensitive Device Operations

Content
View full analysis
bool: """Check if device is online via HTTP whistle""" try: async with aiohttp.ClientSession(timeout=aiohttp.ClientTimeout(total=timeout)) as session: async with session.get(f"http://{ip}/whistle") as resp: return resp.status == 200 except Exception: return False async def get_device_info_http(self, ip: str) -> Optional[dict]: """Get device info via HTTP""" try: async with aiohttp.ClientSession(timeout=HTTP_TIMEOUT) as session: async with session.get(f"http://{ip}/deviceInfo") as resp: if resp.status == 200: return await resp.json() except Exception as e: print(f"❌ HTTP deviceInfo failed: {e}") return None ``` ```python async with aiohttp.ClientSession(timeout=aiohttp.ClientTimeout(total=60)) as session: form = FormData() form.add_field("image", image_data, filename=filename, content_type="image/jpeg") params = { "filename": filename, "gallery": gallery, "show_now": "1" if show_now else "0", } async with session.post(f"http://{ip}/upload", params=params, data=form) as resp: ``` ```python async def get_upstream_settings(self, ip: str) -> Optional[dict]: """Get upstream controller settings""" try: async with aiohttp.ClientSession(timeout=HTTP_TIMEOUT) as session: async with session.get(f"http://{ip}/upstream/pull_settings") as resp: if resp.status == 200: return await resp.json() except Exception as e: print(f"❌ Get upstream settings failed: {e}") return None async def se ...[truncated 3428 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/bloomin8_cli.py:879
Finding

Upstream Token Accepted Through Process Command-Line Arguments

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
SKILL.md:210
Finding

Unpinned Runtime Installation of Third-Party Dependencies

Content
View full analysis
/dev/null || pip install bleak aiohttp pillow ``` The referenced packages are described as follows: ```markdown | Package | Purpose | |---------|---------| | **bleak** | BLE scanning and device communication | | **aiohttp** | HTTP image upload to device | | **pillow** | Image resize/crop to fit screen | ``` ### Technical Analysis The installation command requests the latest available versions of `bleak`, `aiohttp`, and `pillow` without version constraints, hashes, a lock file, a trusted artifact mirror, or an isolated environment. Package installation may execute package-controlled build or installation logic and introduces all transitive dependencies selected by the package resolver at execution time. This makes the effective dependency set mutable after the Skill has been reviewed. A compromised package release, compromised package-index account, malicious transitive dependency, or unexpectedly incompatible future version could affect machines that invoke the documented installation command. The package names themselves are consistent with the imports and declared functionality; no typosquatted package was identified. The finding concerns unsafe and non-reproducible dependency acquisition rather than proof that the named packages are malicious. ### Attack Path 1. The required imports are unavailable on the host. 2. The documented shell expression runs `pip install bleak aiohttp pillow`. 3. `pip` resolves current package and transitive dependency versions from its configured index. 4. A compromised or malicious release is selected, downloaded, and installed. 5. Package build or installation code executes with the permissions of the user running `pip`. 6. Malicious installed code can subsequently execute ...[truncated 659 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (13)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill advertises capabilities that are not actually implemented while omitting others that are implemented, including upstream controller changes and persistent device caching. This mismatch is dangerous because users and orchestrators may authorize the skill under false assumptions, leading to unexpected local-network interaction or configuration changes outside the declared purpose.

Content

No source excerpt is available for this finding.

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · SKILL.md (reported line 65)May include surrounding context.

md
- **GET /open-api/help** — full API docs (no auth)
- **GET /open-api/device/status** — screen dimensions, orientation, interval, pending status (includes `remote_image_on` and `pending_image.remote_image_id` when pending)
- **POST /open-api/push** — push image (URL / file upload / markdown). **Requires `remote_image_on = true`.** Last-wins: only the latest pending image will be displayed.
- **DELETE /open-api/push** — cancel the current pending image. **Requires `remote_image_on = true`.**
- **GET /open-api/push/history** — delivery history with status filter
- **PUT /open-api/device/interval** — change wake frequency (10m / 1h / 3h / 6h). **Requires `remote_image_on = true`.**

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · SKILL.md (reported line 75)May include surrounding context.

md
- **GET /open-api/help** — full API docs (no auth)
- **GET /open-api/device/status** — screen dimensions, orientation, interval, pending status (includes `remote_image_on` and `pending_image.remote_image_id` when pending)
- **POST /open-api/push** — push image (URL / file upload / markdown). **Requires `remote_image_on = true`.** Last-wins: only the latest pending image will be displayed.
- **DELETE /open-api/push** — cancel the current pending image. **Requires `remote_image_on = true`.**
- **GET /open-api/push/history** — delivery history with status filter
- **PUT /open-api/device/interval** — change wake frequency (10m / 1h / 3h / 6h). **Requires `remote_image_on = true`.**

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding

The skill declares broad operational capabilities involving network access and persistent local caching, but does not define any explicit tool scope or permission boundaries. In an agent environment, this increases the chance the agent can perform file and network operations without clear policy gating or informed review, which is a real security issue even if the behavior is partly documented in prose.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 20)May include surrounding context.

md
**Mode selection:** Prefer local mode when available, but always **confirm with the user before performing BLE scans or probing local network hosts**. Suggested flow:
1. Ask the user which mode they prefer (local or cloud), or propose one based on context.
2. If local mode: before scanning, inform the user (e.g. "I'll scan for nearby BLE devices — OK?"). Only proceed with explicit approval.
3. If you have a device IP (from prior `info`, cache, or user-provided), verify reachability (`curl http://<ip>/whistle`) before uploading. Inform the user of the IP being probed.
4. Only fall back to **cloud API** when local is impossible (no BLE hardware, device out of range, IP unreachable after BLE wake).
5. Explicit user intent overrides: "push" / "schedule" → cloud; "upload" / "display now" → local.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
70% confidence
Finding

Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Content

Scanner excerpt · SKILL.md (reported line 191)May include surrounding context.

md
> **Platform notes:**
> - **macOS** — Allow Terminal / your IDE in **System Settings → Privacy & Security → Bluetooth**
> - **Windows** — Requires Windows 10+ (native BLE support via WinRT)
> - **Linux** — Requires BlueZ 5.43+ and DBus (`sudo apt install bluez`); run with a user in the `bluetooth` group or use `sudo`

The CLI script is bundled at `scripts/bloomin8_cli.py` (the **Bloomin8 CLI**) relative to this skill file.

Dynamic Request Target

Medium
Category
Server-Side Request Forgery
Confidence
60% confidence
Finding

Request target host is built from a dynamic or untrusted value. If the host is attacker-influenced, this enables SSRF to arbitrary internal or metadata endpoints.

Content

Scanner excerpt · scripts/bloomin8_cli.py (reported line 320)May include surrounding context.

python
"""Check if device is online via HTTP whistle"""
        try:
            async with aiohttp.ClientSession(timeout=aiohttp.ClientTimeout(total=timeout)) as session:
                async with session.get(f"http://{ip}/whistle") as resp:
                    return resp.status == 200
        except Exception:
            return False

Dynamic Request Target

Medium
Category
Server-Side Request Forgery
Confidence
60% confidence
Finding

Request target host is built from a dynamic or untrusted value. If the host is attacker-influenced, this enables SSRF to arbitrary internal or metadata endpoints.

Content

Scanner excerpt · scripts/bloomin8_cli.py (reported line 329)May include surrounding context.

python
"""Check if device is online via HTTP whistle"""
        try:
            async with aiohttp.ClientSession(timeout=aiohttp.ClientTimeout(total=timeout)) as session:
                async with session.get(f"http://{ip}/whistle") as resp:
                    return resp.status == 200
        except Exception:
            return False

Dynamic Request Target

Medium
Category
Server-Side Request Forgery
Confidence
60% confidence
Finding

Request target host is built from a dynamic or untrusted value. If the host is attacker-influenced, this enables SSRF to arbitrary internal or metadata endpoints.

Content

Scanner excerpt · scripts/bloomin8_cli.py (reported line 445)May include surrounding context.

python
"""Check if device is online via HTTP whistle"""
        try:
            async with aiohttp.ClientSession(timeout=aiohttp.ClientTimeout(total=timeout)) as session:
                async with session.get(f"http://{ip}/whistle") as resp:
                    return resp.status == 200
        except Exception:
            return False

Dynamic Request Target

Medium
Category
Server-Side Request Forgery
Confidence
60% confidence
Finding

Request target host is built from a dynamic or untrusted value. If the host is attacker-influenced, this enables SSRF to arbitrary internal or metadata endpoints.

Content

Scanner excerpt · scripts/bloomin8_cli.py (reported line 429)May include surrounding context.

python
"show_now": "1" if show_now else "0",
                }

                async with session.post(f"http://{ip}/upload", params=params, data=form) as resp:
                    if resp.status == 200:
                        print(f"✅ Image uploaded: {filename}")
                        return True

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The CLI exposes functionality to read and modify upstream controller settings, including endpoint URL, token, and scheduling, which goes beyond simple image delivery and status retrieval. In practice this can redirect the device to a different controller or alter how it pulls content, creating a meaningful configuration-tampering capability that could be abused if this skill is used in a broader agent workflow.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The code allows changing persistent upstream settings and transmitting sensitive values like tokens and controller URLs without any warning, confirmation, or safety interlock. That increases the risk of accidental misconfiguration, credential exposure to unintended endpoints, or silent redirection of the device to an attacker-controlled service.

Content

No source excerpt is available for this finding.

Dynamic Request Target

Medium
Category
Server-Side Request Forgery
Confidence
60% confidence
Finding

Request target host is built from a dynamic or untrusted value. If the host is attacker-influenced, this enables SSRF to arbitrary internal or metadata endpoints.

Content

Scanner excerpt · scripts/bloomin8_cli.py (reported line 477)May include surrounding context.

python
try:
            async with aiohttp.ClientSession(timeout=HTTP_TIMEOUT) as session:
                async with session.put(f"http://{ip}/upstream/pull_settings", json=data) as resp:
                    if resp.status == 200:
                        print("✅ Upstream settings updated")
                        return True